A year in a kitchen, where the feedback loop is about four minutes.

A residency inside an old pub with good bones and an ageing kitchen. Storage was weak, equipment was tired and the mechanical link between the upstairs kitchen and the downstairs dining room did not work, so food and information moved by hand, up and down stairs, all service.

Demand was the harder problem. A quiet weekday could produce almost no covers. A busy weekend could pass 130. A Sunday roast could run lunch through dinner without a gap.

Build a food operation that survives that spread without either running out on Saturday or throwing half the fridge away on Tuesday, and make money doing it.

We started with the physical system, because physical organisation changes behaviour faster than anything you can write down. Deep clean, reorganised prep space, defined what every fridge and freezer was for, built usable dry storage, labelled everything, made cleaning and safety part of the day rather than a folder somebody produces when asked. If an ingredient has no home, someone has to make a decision about it every single time they touch it.

Stock became information. A large whiteboard and simple sheets around the fridges. Deliberately unimpressive, because it had to work at 8pm on a Saturday, not in a process review. What is low, what is turning, what should be frozen, what becomes tomorrow's dish, what should not be reordered.

Then the menu, which is where it gets interesting. Every dish had to answer more than whether it tasted good. What does it cost. How long to prep. How many actions does it add at 130 covers rather than 30. What else uses the same ingredients. How long do those last. Can it be repurposed if the room is quiet. We costed to a 70%-plus gross-margin target while designing around waste, storage, prep time, service load and the standard we wanted to hold. Some dishes disappeared for asking too much of the kitchen mid-service. Others changed because one ingredient could do more useful work elsewhere on the menu.

Then we built the system into the team, because an operation that only works when the same two people are standing in it is not an operation. Prep and service guides, plating photographs, opening and closing routines, daily briefings, pre-service plans, clearer station ownership. People got as much autonomy as they could reliably hold, then more.

The food operation finished approximately 25% ahead of the previous tenancy's revenue, on the historical POS comparison we reviewed at the time.

Within a few months the team could carry services and close the kitchen without us. That was the more useful measure of whether the thing was actually getting stronger.

None of it came from one dramatic restaurant idea. It came from a lot of connected choices becoming less bad at the same time. Layout, stock, prep, menu, training, handoffs, service. Small changes compound when the system connects them.

Service is a good teacher about management style. Something goes wrong, a table is waiting, another ticket lands. You can make the person who made the mistake feel terrible, or you can stabilise the service and work out what happened afterwards. Pressure does not remove accountability. It changes the order in which useful accountability happens.

The bigger lesson was about boundaries. We could improve the kitchen, the food, the team and the service routines. We could not repair the building or change the things decided elsewhere.

That is not unique to hospitality. A good department can sit inside a weak workflow. A strong operating model can still be capped by decisions made somewhere else.

One thing I would do differently. I have a background in positioning and design and I largely left the brand side alone.

I would take both next time.